2023: Week 8

Eight weeks in and we're finally getting reasonable results.  (I didn't post either of the previous two weeks because I got bad results again.)

I used the default values (35 max point difference and a square root curve) and got quick rankings, that resulted in much better stats:

Ranking Stats

Upsets: 473 of 2789 games (16.95948368590893%)
Margins of Victory in upsets: 4823 (8.253191416544029% of total margin points)

This results in the following top 25 rankings:

1. Slippery Rock
2. Shepherd
3. Kutztown
4. California PA
5. East Stroudsburg
6. Indiana PA
7. Assumption
8. Millersville
9. Texas
10. Florida State
11. Ohio State
12. Lock Haven
13. Penn State
14. New Haven
15. West Chester
16. St Anselm
17. Alabama
18. Shippensburg
19. Gannon
20. Oklahoma
21. Bentley
22. Kansas St
23. Washington
24. Oregon State
25. Notre Dame
    

Try it out yourself at: http://kyleburke.info/collegeFootball/cfbRanking.php.  Can you find settings that get better results?

2023: Week 5 (Trying to come back)

I would like to get back to rating teams during the season.  This may be less important now that there's a playoff, but I'm still curious to have some sort of objective basis to see how teams are doing.

Naturally, there are still issues with the algorithm I implemented:

• It doesn't consider overtime.  An overtime win could perhaps be less valuable than a normal win.  I need to work this into the system in some meaningful way.
• It doesn't work if teams play multiple times.  This is a big issue now that conference championships are commonplace throughout NCAA football.  I'm not actually sure what the best way to handle this issue is!
  

Another issue is that I don't have a good mechanism for the case of results that don't converge.   That is the case with what I get in week 5 of this year.  Here are the parameters I chose from the ranker site: http://kyleburke.info/collegeFootball/cfbRanking.php

• 35-point max margin of victory, and
• Using the Quarter-Circle function.

With this I get a back-and-forth semi-tie between Kansas Wesleyan and Southwestern KS.  Instead of converging, it flops back and forth between them.  Currently the algorithm caps out at 18,000 iterations, picking  in an unfair way.  Nevertheless, every week during football season, I wish I were posting these results, so here they go.

The abysmal stats I get for this run are:

Nearly 50% upsets.  Not good.

Here are the top 25 teams from this run.  (You can, of course, check this, but I don't recommend it.)  I expect anyone can pick some parameters that provide better stats.

Anyways, here's the ranking I get:

1. Kansas Wesleyan (they wouldn't have been on top after 18,001 iterations)
2. Southwestern KS
3. Evangel
4. Friends
5. McPherson
6. Ottawa KS
7. St Mary KS
8. Tabor
9. Bethel KS
10. Avila
11. Sterling
12. Bethany KS
13. Georgetown DC
14. Howard
15. Lafayette
16. New Hampshire
17. Norfolk St
18. St Francis PA
19. Abilene Christian
20. Adams St
21. Adrian
22. Air Force
23. Akron
24. Alabama
25. Alabama A&M

The last seven entries aren't in alphabetical order by luck, after St Francis PA, all teams have converged to value zero.  Alabama being in number 24 isn't a good sign that things will rectify in the future.

Nevertheless, I do think things will get better as more weeks go by and more games are played by each team.  I don't know how long it will take before we start to get some meaningful results. 

I'll keep this up as the weeks go by, but I'll also post if I make any improvements.  If you are able to get anything meaningful, please leave a comment!